TRANSACTIONS OF SECTION A. 559 



to August 5 respectively, and minima about the end of April and October 

 respectively. 



Assuming the first order curve to represent the primary solar effect, the 

 purpose of this investigation has been to ascertain the nature and cause of the 

 second order effect. 



Analysis of the temperature at Vienna shows that it does not exist there either 

 to the same extent or at the same epoch. At Agra there is a second order effect 

 of considerable magnitude, but at an entirely different epoch, and hence in no way 

 analogous to the effect in the British Isles. The effect is thus shown to be 

 meteorological and not planetary. 



The effect was first studied for Kew. Its cause was sought in the effects and 

 relative frequency of occurrence of cyclonic and anticyclonic weather. For this 

 purpose the mean temperatures of cyclonic days for each month throughout the 

 year during the five years 1876-1880, and of anticyclonic days during the same 

 period, were separately calculated, and curves were plotted whose ordinates are 

 proportional to the difference between these values and the mean of the ordinates 

 of the first order curve for each month. Both these curves show the main 

 characteristics of the second order curve, and the curve of difference of tempera- 

 ture between cyclonic and anticyclonic weather shows no sign of it. Moreover, 

 by multiplying the percentage of difference of frequency of cyclonic and 

 anticyclonic weather for each month by the difference in temperature, the total 

 effect of type of weather on temperature is obtained, and its curve shows that it 

 does not in any respect resemble the second order effect. It is concluded that 

 although the second order effect has a, meteorological origin the type of the weather 

 plays no part in causing it. 



The effect of wind direction was next examined for the nine-year period 

 1 876-1884. The mean temperatures of the air during the prevalence of barometric 

 gradients towards each of eight points of the compass in each month were 

 separately calculated, and curves of divergence from the first harmonic component 

 were drawn for each wind (taken as being at right angles to the gradient) in the 

 same way as for the cyclonic and anticyclonic curves. Each of these curves shows 

 at least some characteristic of the second order curve ; but on summing them all 

 together a curve is obtained which differs somewhat from the total curve of 

 divergence from first order curve values. 



The effect is largely accounted for as the combined effect of the seasonal 

 variations in temperature of the several winds, and when this part is eliminated 

 the remainder must be attributed to the relative frequency of winds of different 

 temperature. To show this more clearly the winds were grouped together. The 

 mean temperature divergence of east winds is — 3°"1 F. ; of north-east winds 

 — 4°*0F. ; and of north winds — 3°-o F. These winds were grouped as 'cold' 

 winds. Similarly the north-west and south-east winds, whose mean diver- 

 gences are only — 0°-6 F. and — 0°"7 F. respectively, were grouped as ' temperate ' 

 winds, and the west, south-west, and south winds, whose mean divergences are 

 + 1°'7 F., + 2°'2 F., and -f 2°-5 F. respectively, were grouped as ' warm ' winds. 

 Temperature curves were drawn for each of these groups analogous to the curves 

 for the separate winds. Each curve again shows a general resemblance to the 

 second order curve, but it is noticeable that the October-November minimum is 

 especially prominent in the curve for the temperate winds. The mean frequencies 

 of occurrence of these groups in each month during the nine years were also 

 calculated and expressed as a percentage of the total number of days ; the results 

 were plotted on curves whose ordinates are proportional to these percentage 

 frequencies. 



The frequency curve for ' cold ' winds shows a very remarkable maximum 

 frequency in May and a small maximum in November. 



The frequency curve for warm winds shows minima at these times and 

 maxima in February and August, and the freqviency curve for temperate winds, 

 which become distinctly colder in October-November, shows a very high maximum 

 at the end of October. At that time the temperature of these winds is much below 

 the average relative value, and thus the small maximum of the curve of the cold 



